VoxInk Polish & Translate API

Production-ready voice cleanup and 9-language translation for your agent. Skip 2 weeks of prompt engineering — our endpoints are battle-tested on millions of voice typing minutes.

Get an API Key → See the endpoints
What it does

Two endpoints. Both VoxInk specialties.

We bottled a year of prompt-engineering into two HTTP endpoints. Send raw voice transcripts, get clean text back — or translated into any of nine languages with register-aware tone matching.

Polish /api/v1/polish

Turn rambly voice input into well-formatted prose. Filler words, false starts, mishearings, mixed-language code-switching — all cleaned up while keeping the speaker's tone, register, and meaning intact.

"嗯今天那个我去咖啡店" "今天我去咖啡店。"

Translate /api/v1/translate

Translate between English, Chinese, Thai, Indonesian, Vietnamese, Japanese, Korean, Spanish, and French. Register-aware — Thai ครับ/ค่ะ, Indonesian Pak/Ibu, Japanese keigo — auto-selected from source signals. Custom dictionary for domain terms.

"请帮我处理这件事" "Please help me handle this matter."
Why use VoxInk API

Production-quality, integration-ready.

You could write your own polish prompts and chain a translation model. We've already spent a year doing it on real voice data — these are the reasons partners pick our endpoints over rolling their own.

🎯

Battle-tested prompts

One year of production tuning across millions of voice typing minutes.

60-second integration

One curl command to first response. No SDK, no auth flow, just an API key header.

🌐

9 languages, register-aware

Thai politeness particles, Indonesian formal/casual, Japanese keigo, Korean speech levels — automatic.

💰

Transparent pricing

$0.80 per 1M tokens flat. No separate input/output rates, no tier complexity.

Endpoints & examples

Two endpoints, three SDK examples.

Both endpoints take JSON, return JSON, and authenticate via the X-API-Key header. Switch between Polish and Translate below; each shows the request/response shape plus copy-pasteable examples in Python, Node.js, and cURL.

RequestPOST https://api-cn.voxink.ai/api/v1/polish X-API-Key: vk_xxxxxxxxxxxxxxxx_yyyyyy... Content-Type: application/json { "text": "嗯今天那个我去咖啡店" } Response 200 OK{ "polished_text": "今天我去咖啡店。", "request_id": "req_3f2a8c19b04d6e1a", "usage": { "input_tokens": 230, "output_tokens": 50 }, "latency_ms": 432 }

Sample code

import requests

r = requests.post(
    'https://api-cn.voxink.ai/api/v1/polish',
    headers={'X-API-Key': 'vk_xxx...'},
    json={'text': '嗯今天那个我去咖啡店'},
    timeout=10,
)
r.raise_for_status()
print(r.json()['polished_text'])
const r = await fetch('https://api-cn.voxink.ai/api/v1/polish', {
  method: 'POST',
  headers: {
    'X-API-Key': 'vk_xxx...',
    'Content-Type': 'application/json',
  },
  body: JSON.stringify({ text: '嗯今天那个我去咖啡店' }),
});
if (!r.ok) throw new Error(`polish ${r.status}`);
console.log((await r.json()).polished_text);
curl -X POST https://api-cn.voxink.ai/api/v1/polish \
  -H "X-API-Key: vk_xxx..." \
  -H "Content-Type: application/json" \
  -d '{"text": "嗯今天那个我去咖啡店"}'
RequestPOST https://api-cn.voxink.ai/api/v1/translate X-API-Key: vk_xxxxxxxxxxxxxxxx_yyyyyy... Content-Type: application/json { "text": "请帮我处理这件事", "target_language": "English", "dictionary": ["处理=handle"] } Response 200 OK{ "translated_text": "Please help me handle this matter.", "target_language": "English", "request_id": "req_8c0d2fa1e93b7c64", "usage": { "input_tokens": 280, "output_tokens": 60 }, "latency_ms": 587 }

Sample code

import requests

r = requests.post(
    'https://api-cn.voxink.ai/api/v1/translate',
    headers={'X-API-Key': 'vk_xxx...'},
    json={
        'text': '请帮我处理这件事',
        'target_language': 'English',
        'dictionary': ['处理=handle'],
    },
    timeout=10,
)
print(r.json()['translated_text'])
const r = await fetch('https://api-cn.voxink.ai/api/v1/translate', {
  method: 'POST',
  headers: {
    'X-API-Key': 'vk_xxx...',
    'Content-Type': 'application/json',
  },
  body: JSON.stringify({
    text: '请帮我处理这件事',
    target_language: 'English',
    dictionary: ['处理=handle'],
  }),
});
console.log((await r.json()).translated_text);
curl -X POST https://api-cn.voxink.ai/api/v1/translate \
  -H "X-API-Key: vk_xxx..." \
  -H "Content-Type: application/json" \
  -d '{"text":"请帮我处理这件事","target_language":"English","dictionary":["处理=handle"]}'
Pricing

Flat $0.80 per 1M tokens. That's it.

Input + output combined — no separate rates, no tiered surprises. We bill on Gemini's actual usageMetadata counts, so what you pay matches what you consumed.

$0.80 / 1M tokens

Input + output combined. Prepaid balance, deducted per call.

Per-call examples

Use caseTokensCost
Polish (typical)1,000$0.0008
Translate (typical)1,000$0.0008
Long polish (2k in + 500 out)2,500$0.0020

Voice agent usage (per minute of speech)

PatternCalls / minCost / min
Polish only2$0.0016
Polish + Translate4$0.0032
Hour of voice (heavy)240$0.192

Need 10M+ tokens/month? Ask about volume pricing →

Ready to get started?

Contact us privately — we'll issue you an API key after a brief discussion to understand your use case. No signup form, no waitlist, just a real conversation.

📧
partners@voxink.ai
Reply within 24 hours
💬
WeChat QR
Scan to chat on WeChat